OpenModelHub

OpenModelHub is an open-source machine learning model and dataset management platform, inspired by HuggingFace. This project provides a complete solution for storing, managing, and sharing machine learning models and datasets.

🌟 Key Features

• Database Management: Supports basic CRUD operations for models and datasets
• Intelligent Search: Integrated LLM agent supporting natural language queries
• Security Mechanisms: Implements hybrid encryption, access control, and audit trails
• User Interface: Beautiful Streamlit interface with different permissions for users and administrators
• Data Analytics: Provides data insights and visualizations
• File Management: Supports model and dataset upload/download

🚀 Quick Start

Prerequisites

• Python 3.x
• Go 1.x (for security module)
• MySQL 8.x
• Other dependencies listed in requirements.txt

Installation

1. Clone the repository:

git clone https://github.com/yourusername/OpenModelHub.git
cd OpenModelHub

2. Install Python dependencies:

pip install -r requirements.txt

3. Configure environment variables: Create a .env file in the project root directory with the following content:

# -----database-----
DB_USERNAME=$your_DB_USERNAME
DB_PASSWORD=$your_DB_PASSWORD
DB_HOST=0.0.0.0
DB_PORT=3306
TARGET_DB=openmodelhub

# -----agent----
API_KEY=$your_api_key
BASE_URL=$your_base_url

Running the Project

1. Using the startup script (recommended):

python start.py

This will automatically:

• Initialize the database
• Start the security service
• Launch the web interface

2. Manual startup:

# 1. Initialize database
python database/load_data.py
# Choose demo.json for initialization

# 2. Start security service
cd security
go run main.go

# 3. Launch web interface
streamlit run frontend/app.py

👥 User Guide

Login System

• Admin account: username: admin, password: admin
• Regular users: Can register new accounts, default password is admin

Main Features

1. Model Management

   • Browse model repository
   • Upload new models
   • Download models
   • LLM-assisted search

2. Dataset Management

   • Browse datasets
   • Upload new datasets
   • Download datasets
   • Dataset analysis

3. User Management (Admin only)

   • View user list
   • Create/edit users
   • Permission management

4. Data Insights (Admin only)

   • Model analysis
   • Dataset analysis
   • User analysis

🔒 Security Features

• Hybrid encryption (AES-CBC + RSA)
• Argon2Key password hashing
• Role-based access control
• Secure data sharing mechanism
• Comprehensive audit logging

🏗️ Project Structure

OpenModelHub/
├── database/           # Database related code
│   ├── records/       # Sample data
│   └── schema/        # Database schema definitions
├── frontend/          # Streamlit frontend
├── security/          # Go security module
├── doc/              # Documentation
└── requirements.txt   # Python dependencies

📊 Database Design

The project uses a relational database with the following main entities:

• Model
• Dataset
• User
• Affiliation
• Task

For detailed database schema, please refer to the Database Design Documentation

🤝 Contributing

Issues and Pull Requests are welcome! Before submitting a PR, please ensure:

1. Code follows project standards
2. Necessary tests are added
3. Documentation is updated

📝 License

This project is licensed under the MIT License - see the LICENSE file for details

🙏 Acknowledgments

• HuggingFace - Project inspiration
• All project contributors

📚 References

• Argon2: Next Generation Password Hashing Algorithm
• HMAC: Keyed-Hashing for Message Authentication
• RSA: Public-Key Cryptosystem
• More references can be found in the Project Report